Date of Award
Master of Science
This research studied whether computer-generated cloze items using natural language processing methods could promote learning and comprehension of science texts compared to human and random cloze items. Participants recruited from Amazon Mechanical Turk (N = 562) took a pretest on one of three science topics and then read a text on it. Participants then practiced cloze items about the text generated either by a computer (machine), human, or randomly. Cloze items were presented using the MoFaCTS adaptive practice system. After 24 hours participants took a post-test on the text. ANOVA showed a significant effect of cloze type on gain score, and pairwise comparisons found the human conditions had higher gain scores than machine or random conditions. A separate ANOVA on the circulatory system text showed machine had higher gain scores than random. Implications of these findings are discussed.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Whaley, Davis, "Improving Reading Comprehension Of Science Texts With Computer Generated Cloze Item Practice" (2019). Electronic Theses and Dissertations. 1975.